Contact Patterns for HCWs: Not Everyone is the “Average”

Background: To understand how infections spread within hospitals one must have an understanding of how healthcare workers move and interact. Contact-network epidemiology models study populations as networks of potentially disease-spreading interactions between individuals. The challenge in contact network epidemiology is to obtain realistic data from which to generate accurate contact networks.

Objective: We pilot a method for measuring contacts between health-care workers (HCW) using lightweight, non-RFID devices (i.e., motes) that record radio transmissions from other devices in onboard memory.  Using this data we characterize the contact patterns of HCWs in a 21-bed intensive care unit.

Methods: For one week we collected mote-to-mote communication data from devices worn by nearly all HCWs in the Medical Intensive Care Unit at the University of Iowa Hospitals and Clinics.  Motes were separated into three groups corresponding to job roles (nurses, physicians, and critical care support personnel such as pharmacists,respiratory techs, etc.) and distributed in shifts (days from 7am until 7pm, and nights from 7pm until 7am). To protect privacy, motes were randomly assigned within job roles each shift, thus we examine each shift independently of the others.  Messages between motes contain a timestamp, transmitting mote id, receiving mote id, and signal strength. The signal strength of messages transmitted from one mote to another is used as a proxy for distance; all messages below a minimum signal strength corresponding to a distance greater than 6 feet (determined by calibration with ground truth data) are discarded.  We define a continuous contact between two motes to be a time interval where both motes receive at least one message from the other mote once every 30 seconds.  Contacts of duration less than five minutes are also discarded.  All results reported here use t-tests to compare sets of values, with the usual significance threshold of p<0.05.

Results: The length of contacts between pairs of HCWs is not exponentially bounded (i.e. is heavy-tailed), with most contacts being of relatively short duration and a few being of much longer duration. The amount of contact a HCW has in a given shift, measured by total time in contact with another HCW, is also heavy-tailed. This property is conserved even when limited to HCWs of the same job type.  For example, during the night shift, nurses have more contact with other nurses than during the day, with almost five times more contact with other nurses than with physicians, while physicians in the day shift have more than ten times more contact with other physicians than nurses. 

Conclusions: Our results suggest that while HCWs of different job types have different patterns of contact, the contact patterns also vary substantially by individual.  This emphasizes the need for fine-grained contact data for the purpose of modeling disease spread and estimating the effects of infection control interventions in healthcare settings.